Transpiler to support advanced functionality #824
Conversation
dag_drawer: add edge labels and better write register indexing
Clarify dagcircuit test
qiskit/transpiler/_basepasses.py
Outdated
| if isinstance(value, Hashable): | ||
| arguments.append((name, type(value), value)) | ||
| else: | ||
| arguments.append((name, type(value), str(value))) |
There was a problem hiding this comment.
Non-blocking but perhaps we should raise and enforce you use hashable parameters. Don't know.
There was a problem hiding this comment.
I think we should allow non-hashable params, at least for now. Chanching str to repr in 400be85
qiskit/transpiler/_basepasses.py
Outdated
| **kwargs: kwargs to normalize | ||
|
|
||
| Returns: | ||
| Tuple: normalized (args, kwargs) |
There was a problem hiding this comment.
Non-blocking, can we indicate the type (list and dict) of args and kwargs?
qiskit/transpiler/_passmanager.py
Outdated
| 'max_iteration': max_iteration} | ||
|
|
||
| # Default controllers | ||
| FlowController.add_flow_controller('condition', ConditionalController) |
There was a problem hiding this comment.
Should not be module-level code? It is weird to register and re-register each time there we instantiate a new pass manage.
qiskit/transpiler/_passmanager.py
Outdated
| super().__init__(passes, options, **partial_controller) | ||
|
|
||
| def __iter__(self): | ||
| iteration = 0 |
There was a problem hiding this comment.
Non-blocking although I think this is more readable as follows (at least, it avoids the infinite look and make it clear reaching the max_iteration is something bad):
for iteration in range(self.max_iteration):
for pass_ in self.passes
yield pass_
if not self.do_while():
return
raise TranspileError(...)There was a problem hiding this comment.
Much better. Fixed in 4e7d862
|
@1ucian0, defore merging, can we squash the 111 commits into 1 and provide a meaningful description for the commit? |
|
"meaningful description": |
|
Merged in #1060 |
Fix #825
Summary
The current transpiler is very basic and does not implement several of the discussed ideas. This PR tries to be an implementation of them. See details in #825